The fragility of the monetary-energy security nexus: revisiting the ASEAN evidence

1 Introduction

Energy security, commonly defined by the International Energy Agency as the uninterrupted availability of energy sources at an affordable price (International Energy Agency, 2021), is a critical strategic concern for emerging economies. Nowhere is this imperative more acute than in Southeast Asia, where primary energy demand has grown by approximately 3% per annum over the past two decades, consistently outpacing domestic supply (BP, 2023). The ASEAN Centre for Energy's eighth ASEAN Energy Outlook (AEO8) projects a widening energy trade deficit through 2050 under current policy trajectories (ACE, 2024), while the ASEAN Investment Report 2025 documents that annual investment into the region's clean-energy systems still captures only about 20% of the levels required to meet the region's net-zero pathway, leaving a substantial gap in new generation, grid modernization, and storage capacity (ASEAN Secretariat and UNCTAD, 2025). These diverging supply, demand, and investment trajectories underscore the urgency of understanding all macroeconomic drivers, including monetary policy, that influence energy vulnerability.

The energy uncertainty confronting ASEAN is not merely a supply-side problem; it encompasses price volatility, geopolitical supply disruption, and the institutional capacity to manage these risks. Building on this multidimensional view, Panda and Mahalik (2026) demonstrate that renewable energy generation and governance quality jointly mitigate energy uncertainty in emerging economies. However, governance capacity alone does not fully account for the macroeconomic transmission of energy shocks. Monetary policy conditions also shape how energy uncertainty propagates across economies with different trade positions. Figure 1 illustrates the co-movement between macroeconomic conditions and energy sector uncertainty across ASEAN, revealing that periods of monetary tightening coincide with heightened energy import volatility in deficit economies but not in surplus ones. This graphical evidence motivates the formal econometric investigation that follows.

While energy security has traditionally been analyzed through the lens of energy-specific instruments, the potential influence of broader macroeconomic policies remains insufficiently explored in the ASEAN context. Monetary policy, through interest-rate adjustments and their effects on credit, investment, exchange rates, and aggregate consumption, can exert indirect effects on energy security through multiple channels. The closest antecedent to our work is Wang et al. (2022), who examine the monetary policy-energy security nexus for 16 Asian countries using System GMM. Their contribution established the empirical relevance of this research question but faces fundamental identification limitations in the macro-panel context where the cross-sectional dimension (N) is small relative to the time dimension (T): System GMM relies on large-N asymptotics for the consistency of its internal instruments, and it imposes homogeneous long-run slope coefficients across countries.

Several structural features distinguish ASEAN from the broader Asian panel examined in Wang et al. (2022) and justify a dedicated investigation. First, ASEAN economies display extreme heterogeneity in energy endowments: Brunei and Malaysia are net energy exporters, while the Philippines, Singapore, Thailand, and Vietnam are heavily dependent on imported hydrocarbons. This within-region disparity generates fundamentally different monetary transmission pathways that are masked when ASEAN is pooled with East Asian and South Asian economies. Second, exchange-rate regimes across ASEAN range from currency boards (Brunei) to managed floats (Indonesia, Thailand) and tightly managed regimes (Vietnam), implying that the exchange-rate channel of monetary transmission operates with different intensities across the region. Third, several ASEAN economies maintain extensive government intervention in energy pricing through fuel subsidies, which attenuates the demand-side channel of monetary transmission in ways that are less prevalent in the advanced Asian economies. Fourth, ASEAN capital markets remain relatively shallow compared to those in Japan, South Korea, or China, implying that the investment channel through which interest rates affect renewable energy deployment is mediated by different financial structures. These features collectively mean that empirical findings from a broader Asian sample cannot be directly extrapolated to ASEAN, and that an ASEAN-specific analysis is needed to inform regional energy and monetary policy coordination.

Because of these restrictions, we adopt the Pooled Mean Group (PMG) estimator as our baseline specification. PMG is purpose-built for the (small N, moderate T) configuration that characterizes ASEAN macroeconomic panels: it dispenses with the large-N asymptotic arguments that System GMM cannot defend at N = 7–11, and it nests the analysis within a panel error-correction architecture that separates long-run equilibrium relationships from country-specific short-run dynamics. To be clear, PMG is not assumption-free. Like System GMM, it imposes long-run homogeneity across countries, but with two crucial differences. First, this restriction enters PMG as an explicit modeling choice that can be formally tested against the less restrictive Mean Group (MG) alternative via the Hausman statistic, rather than being bundled silently with an instrumentation strategy. Second, the homogeneity constraint is imposed only on the long-run equilibrium; short-run dynamics, error variances, and the speed of adjustment to equilibrium are allowed to vary freely across the seven ASEAN economies. The result is a framework in which the central modeling assumption is transparent, testable, and considerably narrower in scope than the bundle of restrictions required by System GMM in this small-N setting.

This paper makes four methodological contributions. First, we extract exogenous monetary policy shocks (MonShock) via country-level Structural VAR models in which the real effective exchange rate is included as an endogenous variable, making the identification robust to the external-balance channel emphasized by the open-economy macro literature. Second, we adopt a panel time-series framework (PMG with MG and CS-ARDL as robustness checks) appropriate for macro-panels of the dimensions available in ASEAN. Third, we replace a statistically fragile split-sample analysis with a pooled PMG specification augmented by a MonShock × DeficitDummy interaction term, which preserves degrees of freedom while still capturing heterogeneity between energy-surplus and energy-deficit economies. Fourth, we implement second-generation CIPS unit-root tests and Pesaran CD statistics to address cross-sectional dependence in the panel and apply an inverse hyperbolic sine (IHS) transformation to the dependent variable to mitigate the distorting influence of Brunei's extreme negative net-import observations.

Our results deliver a more conservative message than the prior literature. After applying these corrections, the long-run coefficient of MonShock on energy import dependence and its interaction with the deficit dummy are not statistically significant at conventional levels under PMG with robust standard errors. The same pattern holds under MG and CS-ARDL. The renewable energy share does not respond to monetary shocks under any specification. Taken together, these results suggest that the apparent security gains from contractionary monetary policy documented in earlier work are fragile to methodological choices. Once we correct for outliers, add subsidy controls, allow the exchange rate to respond endogenously, and use estimators suited to the panel structure of our data, the effect disappears. This fragility means that the earlier findings should not be read as a stable feature of how monetary policy works in ASEAN.

The remainder of this paper is organized as follows. Section 2 reviews the literature on energy security and its monetary determinants and develops the theoretical framework. Section 3 details the empirical strategy. Section 4 presents the results. Section 5 discusses implications and limitations, and Section 6 concludes.

2 Literature review

2.1 Conceptualizing and measuring energy security

Energy security is a multi-dimensional concept that has been defined and measured in a variety of ways. The Asia Pacific Energy Research Centre's influential “4-As” framework identifies availability, accessibility, affordability, and acceptability as four interrelated pillars of a secure energy system (Ziaei, 2018). Cherp and Jewell (2014) criticize this framework as incomplete and arbitrary, arguing for a more rigorous treatment grounded in the three analytical perspectives of sovereignty, robustness, and resilience. Asafo et al. (2024) subsequently proposed an “energy humanization” index that adapts the 4-As framework to five dimensions (availability, affordability, acceptability, technology, and usage), and apply it to Ghanaian data. Their central empirical finding carries direct implications for emerging economies: while the availability dimension improved steadily over time, both affordability and the uptake of modern energy technologies deteriorated, suggesting that aggregate-level security indicators can mask regressive distributional and technological dynamics.

Kharazishvili et al. (2025) develop these concerns into a “sustainable development trajectory” framework for energy security, using Ukraine as a case study. In their formulation, energy security is not assessed against static thresholds but against a dynamic trajectory that co-evolves with economic, environmental, and institutional conditions. Panda and Mahalik (2026) extend this line of inquiry to emerging economies, demonstrating that renewable energy generation and governance quality jointly mitigate energy uncertainty, and that the effectiveness of these mitigating factors varies with institutional capacity. Their finding that governance quality moderates the energy uncertainty–renewable energy relationship is particularly relevant to the ASEAN context, where institutional heterogeneity across member states is substantial. This perspective is directly relevant to the ASEAN case: countries like Indonesia and Vietnam are changing rapidly, both in their energy mix and in the institutions that govern it. For such countries, a single snapshot of import dependence or HHI score misses the bigger question: is the country actually moving in a sustainable direction? Our study accommodates this insight by focusing on the long-run equilibrium coefficients of a panel ECM rather than on snapshot correlations, and by complementing the import-dependence metric with a renewable energy share indicator that captures directional movement along the energy transition.

Empirical work in this area has converged on several quantitative metrics: the energy self-sufficiency ratio, the energy import dependency ratio, the Herfindahl-Hirschman Index (HHI) of supply diversification, and the Shannon index (Yurnaidi et al., 2024). In this paper we focus on net energy import dependence and renewable energy share, both drawn from standardized international sources, while acknowledging that no single metric captures all dimensions of the concept.

2.2 Theoretical framework: monetary transmission channels to energy security

The theoretical foundation for examining the monetary policy–energy security nexus rests on three distinct transmission channels, each with different implications for ASEAN economies. We develop these channels here to provide an explicit theoretical basis for the empirical specifications that follow.

2.2.1 The demand channel

Standard New Keynesian theory predicts that contractionary monetary policy reduces aggregate demand through higher borrowing costs and reduced credit availability (Galí, 2015). In the energy context, this demand contraction reduces energy consumption and, for net importers, lowers energy import dependence. The strength of this channel depends on the interest-rate sensitivity of aggregate demand, which in turn depends on financial market depth, the share of bank-intermediated credit, and the prevalence of informal economic activity. In ASEAN, where informal sectors remain large (particularly in Vietnam, Indonesia, and the Philippines) and where government-administered energy prices attenuate the price signal, the demand channel is expected to be weaker than in advanced economies. Fuel subsidies further blunt the demand response by insulating consumers from the full cost of energy, reducing the elasticity of energy demand with respect to macroeconomic conditions.

2.2.2 The investment channel

Higher policy rates raise the weighted average cost of capital (WACC) for energy projects, with disproportionate effects on capital-intensive renewable energy investments that have high upfront costs but low marginal operating costs (Egli et al., 2018). This channel operates asymmetrically across energy technologies: fossil-fuel projects, which have lower capital intensity relative to operating costs, are less sensitive to interest-rate changes than wind and solar installations (Serebriakova et al., 2026). For ASEAN economies seeking to diversify away from imported fossil fuels through domestic renewable capacity, this asymmetry creates a tension between monetary stabilization and energy transition objectives. The channel is further mediated by the depth and sophistication of domestic capital markets; in economies with shallow bond markets and limited project-finance infrastructure, the investment channel may operate primarily through bank lending rates rather than through capital market pricing.

2.2.3 The exchange-rate channel

In open economies, contractionary monetary policy typically appreciates the domestic currency, reducing the local-currency cost of energy imports and potentially increasing import dependence by making imported energy relatively cheaper. This channel is particularly relevant for ASEAN economies with managed-float exchange-rate regimes, where monetary policy actions have direct exchange-rate consequences. The direction of the net effect on energy security depends on whether the demand-reducing effect of higher rates dominates the import-cheapening effect of currency appreciation. In ASEAN, where exchange-rate management is a core monetary policy objective for several central banks, the exchange-rate channel cannot be treated as a residual; it must be modeled explicitly, which motivates our inclusion of REER as an endogenous variable in the SVAR identification (Section 3.2).

These three channels generate an ambiguous net prediction for the effect of monetary policy on energy security, with the sign depending on the relative strength of each channel and the structural characteristics of each economy. This theoretical ambiguity provides the analytical motivation for our empirical investigation and explains why the estimated effects may differ across energy-surplus and energy-deficit economies within ASEAN.

2.3 Monetary policy and the energy sector: demand, investment, and technology-bias channels

Two competing transmission channels dominate the theoretical literature. The demand channel operates through aggregate activity: contractionary monetary policy dampens economic growth, thereby reducing energy consumption and import requirements (Miranda-Pinto et al., 2023; Ziaei, 2018). High-frequency studies confirm a rapid price dimension of this channel, with an unexpected 100 basis-point increase in the US federal funds rate leading to immediate declines in crude oil prices of 2–3% (Frankel and Hardouvelis, 1983; Rosa, 2014). More recent evidence reinforces and extends these foundational findings. Miranda-Pinto et al. (2023) show that US monetary shocks transmit to commodity prices through both demand and financial channels, with the effect on energy commodities being particularly pronounced. Aliyev and Kočenda (2023) document that ECB monetary policy surprises significantly affect commodity prices, including energy, through portfolio rebalancing and exchange-rate channels. Bastianin and Manera (2018) provide updated evidence that monetary policy surprises generate statistically significant and economically meaningful responses in oil prices within the first month, with the magnitude varying across conventional and unconventional policy regimes. These studies collectively establish that the high-frequency price channel remains operative across different monetary regimes and time periods.

The investment channel, by contrast, operates through the cost of capital: higher policy rates raise required returns and can deter new energy investments, particularly in capital-intensive renewable projects (Arkolakis and Walsh, 2024; Egli et al., 2018; Hilmarsson, 2017). Detemple et al. (2024) show that under stochastic interest rate and regime-switching volatility, the sign of the investment-channel effect can be non-monotonic.

Recent work sharpens this investment channel into a specific concern about technology bias in monetary policy. Serebriakova et al. (2026), using a fixed-effects panel IV estimator on 28 European countries and 10 energy technologies over 2001–2024, document a striking asymmetry: a 25 basis-point increase in ECB policy rates is associated with approximately a 3.2% reduction in onshore wind installed capacity and a 5.3% reduction in solar PV capacity, while fossil fuel, hydropower, and nuclear installations are statistically unaffected. Their results confirm that monetary tightening affects renewable and fossil-fuel technologies very differently. Higher interest rates raise borrowing costs directly, and at the same time make it harder for capital-intensive firms, including most wind and solar developers, to access credit. This asymmetry matters for energy security: because wind and solar account for most new capacity additions, anything that slows their deployment also slows the pace at which importing countries can replace imported fossil fuels with domestic renewables.

Complementary studies examine monetary-uncertainty and firm-level transmission. Sohail et al. (2021) show via NARDL models that monetary policy uncertainty has asymmetric effects on renewable vs. non-renewable energy consumption in the US. Akan (2023) uses structural equation modeling to establish renewable energy as a mediator between monetary conditions and carbon emissions. Dagar et al. (2024) find, for 26 countries across three continents, that heightened economic policy uncertainty narrows the spectrum of energy sources exploited and thereby reduces diversification. Rao et al. (2024) use quantile connectedness methods at the firm level for five Asian economies and show that renewable energy firms are significantly more vulnerable to interest-rate volatility than fossil fuel firms, with the disparity concentrated among highly leveraged companies. Hashmi et al. (2022) apply a Fourier-augmented ARDL to US data and find that expansionary monetary policy promotes renewable energy consumption in both the short and long run.

Among existing empirical studies, Wang et al. (2022) is the most directly comparable to ours. Using System GMM on a panel of 16 Asian countries, they provide the first direct evidence that standard monetary policy instruments shape energy security outcomes. They also document that this relationship became stronger after the 2008 global financial crisis. Their paper established a methodological template that later studies have adopted, but as discussed in the Introduction, the combination of a small cross-sectional dimension, the assumption of homogeneous long-run coefficients, and the use of internal instruments raises concerns about instrument proliferation and small-sample bias that we address directly in our estimation strategy.

2.4 Research gaps and contributions

Three gaps persist in this literature. First, existing empirical evidence is drawn predominantly from advanced economies, where monetary transmission operates through mature financial markets and stable institutions (Aliyev and Kočenda, 2023; Rosa, 2014). Southeast Asian economies are characterized by shallow capital markets, heterogeneous exchange-rate regimes, large informal sectors, and starkly heterogeneous energy endowments, none of which allow direct extrapolation from developed-economy results. Second, the short-run and long-run components of monetary transmission to energy security have not been formally decomposed in an ASEAN context. Third, the role of structural heterogeneity between energy-surplus and energy-deficit economies within ASEAN has not been empirically tested in a framework that preserves statistical power. This study addresses all three gaps by (i) extracting exogenous monetary policy shocks via country-level structural VAR models that include the real effective exchange rate as an endogenous variable, (ii) applying PMG panel ECM estimators appropriate for macro-panels, complemented by MG, CS-ARDL, and panel VAR as robustness checks, and (iii) introducing a MonShock × DeficitDummy interaction term in place of statistically underpowered split-sample estimation.

3 Methodology

3.1 Data and variables

We analyze an annual panel of seven ASEAN member states: Brunei Darussalam (BRN), Indonesia (IDN), Malaysia (MYS), the Philippines (PHL), Singapore (SGP), Thailand (THA), and Vietnam (VNM) over 2000–2023. Three ASEAN members are excluded due to data limitations: Cambodia (no continuous interest rate series), Lao PDR (insufficient time-series length for SVAR estimation), and Myanmar (missing REER and truncated post-2019 macroeconomic series). Table 1 defines all variables and reports their sources.

Table 1. Variable definitions and sources.

TPES, total primary energy supply. The RenewShare series excludes traditional biomass (fuelwood, charcoal, unmodernised agricultural residues) to focus on policy-relevant modern renewable capacity additions.

Net energy import dependence (ImportDep) is drawn from the World Bank's World Development Indicators and is defined as net energy imports as a share of total energy use. To mitigate the distortionary influence of Brunei's extreme negative outliers, which arise from Brunei's position as a major hydrocarbon exporter whose reported “net imports / use” ratio can exceed−800% in individual years, we transform ImportDep with the inverse hyperbolic sine (IHS) function, ImportDep_IHS = arcsinh(ImportDep) = ln(ImportDep + √(ImportDep² + 1)). The IHS transformation is monotonic and symmetric around zero, behaves like a log transformation for large positive or negative values, is defined at zero and for negative numbers, and compresses the scale sufficiently to prevent a single country from dominating the pooled estimates while preserving the sign of net-export vs. net-import status.

The renewable energy share (RenewShare) is constructed from IEA energy balances and IRENA capacity statistics. Following the IEA convention, our RenewShare series includes modern biomass and waste but excludes traditional use of solid biomass (fuelwood, charcoal, agricultural residues used without modern conversion technologies). This exclusion is important for interpretation: traditional biomass is a large share of household energy use in some ASEAN economies (notably Cambodia and Myanmar, which are outside our sample, and parts of Indonesia and the Philippines) and its inclusion would inflate the RenewShare series without reflecting policy-relevant renewable-capacity additions. Our RenewShare is therefore best interpreted as a measure of the modern renewable energy transition rather than of total non-fossil energy use.

Our key explanatory variable is MonShock, an SVAR-identified exogenous monetary policy shock described in Section 3.2. The augmented control vector also includes the real effective exchange rate (REER), a US-ASEAN Fed-rate spread (FedSpread), the natural log of real GDP per capita (lnGDPpc), and the natural log of the Brent crude oil price (lnOil). Finally, we include a binary SubsidyProxy control equal to one for Indonesia in 2000–2014 and for Malaysia in 2000–2013, capturing the two largest documented ASEAN fuel-subsidy episodes within the sample period. Indonesia removed most of its gasoline subsidy in 2015 following a peak subsidy bill exceeding 4% of GDP in 2012; Malaysia began phasing out its RON95 subsidy in 2014. Because comprehensive country-year fuel subsidy series are not available for the entire ASEAN-7 sample over 2000–2023, binary proxy is a conservative approximation of subsidy-regime heterogeneity; the limitation is discussed in Section 5.5.

3.2 Identification of monetary policy shocks via SVAR

For each country, we estimate a five-variable structural VAR over the longest continuous data segment available, ordered as [lnOil, GDP growth, CPI, REER, IR]. This extends the standard four-variable Cholesky identification in Christiano et al. (1999) by inserting REER before IR, reflecting the view that in small open ASEAN economies the monetary authority observes the exchange rate before setting the policy rate but the exchange rate does not react contemporaneously to a domestic policy shock. Making REER endogenous to the identification system is essential to avoid attributing exchange-rate-driven pass-through of commodity prices to monetary shocks, a concern emphasized in the open-economy monetary-transmission literature. The lag length is restricted to p = 1 to preserve degrees of freedom in samples of roughly 20 observations per country. This restriction can be read as a frequentist analog of a Bayesian VAR with a tight Minnesota prior around p = 1. MonShock is the structural residual of the interest rate equation, standardized across the pooled sample to have zero mean and unit variance.

3.3 Panel diagnostic sequence

Before selecting the estimation method, we conduct a systematic panel diagnostic sequence following the recommended protocol in the heterogeneous panel literature (Chudik and Pesaran, 2015; Panda et al., 2025). The diagnostic sequence proceeds in four steps: (i) cross-sectional dependence testing, (ii) slope heterogeneity testing, (iii) unit root testing under cross-sectional dependence, and (iv) cointegration testing. The results of these diagnostics determine the appropriate estimation strategy.

Step 1: Cross-sectional dependence. We apply the Pesaran (2021) CD test to all level variables. The results (reported in Table 3) decisively reject cross-sectional independence for all variables, with CD statistics ranging from 3.0 to 9.7 (all p < 0.01). This finding rules out first-generation panel estimators that assume cross-sectional independence and motivates the use of second-generation unit-root tests (CIPS) and the CS-ARDL robustness specification.

Step 2: Slope heterogeneity. We implement the Pesaran and Yamagata (2008) test for slope heterogeneity. The adjusted test statistic is 4.87 (p < 0.01), decisively rejecting the null of homogeneous slopes across the seven ASEAN economies. This result implies that imposing fully homogeneous coefficients, as System GMM does, is not supported by the data. The PMG estimator accommodates this finding by allowing heterogeneous short-run coefficients while maintaining homogeneous long-run coefficients, a restriction that can be tested against the fully heterogeneous MG alternative. The CS-ARDL further accommodates both slope heterogeneity and cross-sectional dependence.

Step 3: Unit root tests. Given the presence of cross-sectional dependence, we employ the Pesaran (2007) CIPS test rather than first-generation alternatives such as IPS. Results are reported in Table 3 and discussed in Section 4.2. The mixed integration orders [I(0) and I(1)] are fully compatible with the ARDL/ECM framework.

Step 4: Cointegration. We apply the Westerlund (2007) ECM-based panel cointegration tests, which allow for cross-sectional dependence and heterogeneous short-run dynamics. The group-mean statistics (Gτ and Gα) test the null of no cointegration for each country, while the panel statistics (Pτ and Pα) test the null for the panel as a whole. Using bootstrapped p-values (with 800 replications) to account for cross-sectional dependence, the Gτ statistic rejects the null of no cointegration at the 5% level (p = 0.032), while the panel Pτ statistic is marginally significant (p = 0.078). These results provide sufficient evidence of a long-run equilibrium relationship to justify the error-correction specification, though the relatively weak cointegration signal is consistent with the slow error-correction speeds documented in Section 4.

The diagnostic results collectively point to a panel characterized by cross-sectional dependence, slope heterogeneity, mixed integration orders, and weak cointegration. This configuration favors the PMG estimator as the baseline (accommodating mixed I(d) orders and heterogeneous short-run dynamics within an ECM structure) with CS-ARDL as the primary robustness check (additionally accommodating cross-sectional dependence). We retain both MG and Panel VAR as further robustness exercises. The choice of estimator is thus data-driven rather than arbitrary, following the diagnostic protocol recommended by Panda et al. (2025).

3.4 Econometric strategy

The macro-panel structure of our data (N = 7, T up to 23) dictates our choice of estimator. Our baseline is the Pooled Mean Group (PMG) estimator of Pesaran et al. (1999), which constrains long-run coefficients to be homogeneous across countries while allowing heterogeneous short-run dynamics and country-specific error-correction speeds. The Hausman test is used to assess long-run homogeneity against the Mean Group (MG) alternative of Pesaran and Smith (1995), which imposes no cross-country restrictions. As a further robustness check against cross-sectional dependence, we estimate a Cross-Sectionally Augmented ARDL (CS-ARDL) specification following Chudik and Pesaran (2015) in which the ECM regressors are augmented with cross-sectional averages. The Pesaran (2021) CD test motivates the CS-ARDL specification; the Im-Pesaran-Shin first-generation unit-root test is replaced by the second-generation CIPS test of Pesaran (2007), which is robust to cross-sectional dependence.

The baseline error-correction model, written for the IHS-transformed dependent variable, is:

ΔimportDep_IHSit=αi+ϕi(ImportDep_IHSi,t-1-θ′Xit) +δi′ΔXit+εit

where X includes MonShock, MonShock × DeficitDummy, lnGDPpc, lnOil, REER, and SubsidyProxy. Under PMG, θ is restricted to be common across countries and is estimated via concentrated log-likelihood on the pooled panel, while ϕ_*i* and δ_*i* are estimated country-by-country. Standard errors are heteroskedasticity-robust (HC1). The long-run effect for deficit countries is θ_MonShock + θ_Interaction, and for surplus countries it is θ_MonShock alone.

As a further robustness check, we estimate a Panel VAR (PVAR) of order 1 on the within-transformed variables (MonShock, ImportDep_IHS, RenewShare), following a pooled approach with country fixed effects removed by the within transformation. The PVAR does not impose the ARDL long-run restriction and offers an alternative reading of the dynamic interactions among the three key variables. We prefer PVAR over Bayesian SVAR for this robustness exercise because the pooled structure recovers degrees of freedom without requiring the specification of subjective priors, while still respecting the reviewer's concern about over-parameterisation.

We address cross-sectional dependence, which the Pesaran CD test confirms is pervasive in our panel (see Section 4.2), by using the CIPS second-generation unit-root test for I(d) classification and by augmenting the ARDL with cross-sectional averages in the CS-ARDL robustness check. We retain CS-ARDL only as a robustness specification, not as a main result. With T = 23 and N = 7, the cross-sectional averaging absorbs a substantial share of the identifying variation, and the estimator's small-sample properties in this dimension remain uncertain. This is an appropriate use of the method rather than a principal inferential tool in a sample of this size.

4 Results

4.1 Descriptive statistics

Table 2 reports descriptive statistics for the working sample, including the untransformed and IHS-transformed import dependence series. The raw ImportDep series ranges from approximately −840% (Brunei, reflecting its status as a hydrocarbon exporter) to 98% (Singapore, which imports nearly all of its primary energy), with strongly negative skewness and high kurtosis. After IHS transformation, the range compresses to (−7.4, 6.5), skewness declines to approximately zero, and kurtosis falls substantially, confirming that the transformation successfully mitigates the Brunei outlier problem. MonShock is by construction centered on zero with unit variance. The SubsidyProxy binary equals one in 17.3% of observations, corresponding to the 15 Indonesia-years (2000–2014) and 14 Malaysia-years (2000–2013) in the sample.

Table 2. Descriptive statistics (2000–2023, seven ASEAN countries).

ImportDep_IHS = arcsinh(ImportDep). MonShock is the standardized SVAR residual from the interest-rate equation in a five-variable system [lnOil, GDP growth, CPI, REER, IR]. N = 7 countries (BRN, IDN, MYS, PHL, SGP, THA, VNM). Observation counts differ across variables due to country-specific data availability.

4.2 Panel diagnostic results

Table 3a reports the results of the Pesaran and Yamagata (2008) slope heterogeneity test. The adjusted Δ statistic is 4.87 (p < 0.01), strongly rejecting the null hypothesis of homogeneous slope coefficients across the seven ASEAN economies. This result confirms that the short-run dynamics of the monetary policy-energy security relationship differ meaningfully across countries, consistent with the structural heterogeneity described in the Introduction.

Table 3a. Pesaran-Yamagata slope heterogeneity test.

The Pesaran and Yamagata (2008) Δ test is adjusted for small samples. The null hypothesis is that slope coefficients are homogeneous across all seven ASEAN economies.

^***p < 0.01.

Table 3b reports the Pesaran (2007) CIPS test for panel unit roots and the Pesaran (2021) CD test for cross-sectional dependence. CIPS is computed as the average of country-specific CADF t-statistics and its 5% critical value for a panel with N = 7 and T is approximately 20 with a constant is approximately −2.27. The test classifies ImportDep_IHS, RenewShare, REER, and lnOil as I(1) in levels and I(0) in first differences. MonShock, FedSpread, and lnGDPpc are I(0) in levels, consistent with their construction (SVAR residuals and interest-rate differentials are stationary by construction). The mixed integration orders are fully compatible with the ARDL/ECM framework, which does not require all variables to be integrated of the same order. The Pesaran CD test decisively rejects cross-sectional independence for all level variables (CD statistics of 3.0–9.7, p < 0.01), confirming the need for CIPS rather than IPS and motivating the CS-ARDL robustness check.

Table 3b. CIPS panel unit root tests and Pesaran CD tests.

CIPS = Pesaran (2007) second-generation panel unit-root test. CD = Pesaran (2021) cross-sectional dependence test. The 5% critical value for CIPS with constant, N = 7, T ≈ 20 is approximately −2.27. ^***p < 0.01, ^*p < 0.10. lnOil and FedSpread are common across countries and so CD is not reported.

Table 3c reports the Westerlund (2007) ECM-based panel cointegration test results. Using bootstrapped critical values to account for cross-sectional dependence, the group-mean Gτ statistic rejects the null of no cointegration at the 5% level (p = 0.032), while the panel Pτ statistic is marginally significant at the 10% level (p = 0.078). The Gα and Pα statistics do not reject the null. The mixed cointegration evidence is consistent with the interpretation that a long-run equilibrium exists between monetary conditions and energy import dependence, but that the relationship is weak and subject to slow adjustment, as reflected in the error-correction speeds estimated below.

Table 3c. Westerlund (2007) ECM-based panel cointegration tests.

Westerlund (2007) ECM-based panel cointegration test. Gτ and Gα are group-mean statistics; Pτ and Pα are panel statistics. Bootstrap p-values based on 800 replications to account for cross-sectional dependence. The null hypothesis is no cointegration. ^**p < 0.05, ^*p < 0.10.

4.3 Main PMG estimation with interaction

Table 4 reports the main PMG estimation results. For comparison, we also report Mean Group (MG) estimates. As an additional robustness check, we estimate the model with CS-ARDL, which accounts for cross-sectional dependence by including cross-sectional averages of the regressors. Standard errors are heteroskedasticity-robust (HC1); clustering by country with N = 7 yields known small-cluster conservativeness and is reported in Supplementary material.

Table 4. PMG, MG, and CS-ARDL estimation with MonShock × DeficitDummy interaction.

Dependent variable: ImportDep_IHS. Standard errors in parentheses are heteroskedasticity-robust (HC1). The dependent variable is ImportDep_IHS. MonShock is the SVAR-identified monetary policy shock, DeficitDummy = 1 for PHL, SGP, THA, VNM. PMG long-run coefficients are restricted to be homogeneous; MG coefficients are simple averages of country-specific ARDL long-run estimates; CS-ARDL augments the ARDL with cross-sectional averages of ImportDep_IHS, MonShock, lnGDPpc, and REER. The Hausman test of PMG against MG is not reported because the MG variance matrix is not positive-definite in this small-N setting, a known limitation of the Hausman test when N is small and the number of long-run parameters is large; a parameter-by-parameter Wald comparison between PMG and MG point estimates does not reject the PMG restriction for any individual coefficient at the 5% level. The half-life is computed as ln(0.5) / ln(1 + ϕ), which is the number of years required for half of a unit shock to the long-run relationship to dissipate under the estimated error-correction speed.

The central message of Table 4 is the absence of statistical significance for the monetary policy coefficients under robust standard errors. Under PMG, the point estimate of MonShock is 2.502 with a standard error of 2.084 (t = 1.20), and the interaction coefficient MonShock × DeficitDummy is −2.142 with a standard error of 2.080 (t = −1.03). The implied long-run effect of a one-standard-deviation contractionary shock on ImportDep_IHS in deficit economies is therefore 2.502–2.142 = 0.360, economically close to zero, while the effect in surplus economies is 2.502. Neither coefficient, nor their sum, attains conventional levels of statistical significance. The same qualitative pattern holds under MG and CS-ARDL: the point estimates shift noticeably across estimators, but none produces a t-statistic exceeding 1.96 for any MonShock-related coefficient.

The error-correction term is negative and of plausible magnitude under all three estimators. Under PMG, ϕ = −0.077 with t = −1.58, implying a half-life of approximately 8.7 years. Under MG, ϕ averages −0.603 across countries, reflecting the combination of faster-adjusting deficit economies and slower-adjusting surplus economies. The slow pooled adjustment speed under PMG is consistent with the interpretation that ImportDep is a deeply structural variable whose equilibrium is governed by the underlying resource endowment and generation mix, not by short-run monetary conditions.

The point estimate on SubsidyProxy is positive under all three estimators (1.256–3.569) but statistically insignificant. The sign is consistent with the hypothesis that fuel subsidies weaken the demand-side channel of monetary transmission and leave import dependence structurally higher, but the confidence intervals are too wide to support strong inference. We return to this limitation in Section 5.

4.4 Alternative dependent variable: renewable energy share

Table 5 reports PMG estimates for RenewShare as the dependent variable, using the same regressor set and augmented control vector as Table 4. The results are stark: no monetary coefficient, long-run or short-run, direct or interacted, is statistically significant, and all point estimates are economically trivial. The point estimate on MonShock in the long run is 0.002 (SE = 0.011) and on MonShock × DeficitDummy is −0.013 (SE = 0.023). The coefficient on SubsidyProxy is negative (−0.068, SE = 0.066, t = −1.03). The sign is consistent with the hypothesis that fuel subsidies slow the renewable transition, though the estimate is too uncertain to distinguish from zero at conventional significance levels.

Table 5. PMG estimation for the renewable energy share.

Dependent variable is RenewShare (modern renewable energy / TPES, excluding traditional biomass). HC1 robust standard errors in parentheses. Half-life computed as ln(0.5) / ln(1 + ϕ). ^**p < 0.05

Taken together with the ImportDep_IHS results, the RenewShare estimates point to a consistent interpretation: monetary policy in ASEAN does not appear to have a statistically detectable causal effect on either the quantitative or the compositional dimension of energy security over 2000–2023. Any improvement in import dependence that might eventually emerge from contractionary policy works through gradual demand adjustment rather than through structural reallocation toward domestic renewable supply.

4.5 Robustness: Panel VAR

To provide a fully atheoretical complement to the PMG/CS-ARDL estimates, we estimate a Panel VAR of order 1 on the within-transformed variables (MonShock, ImportDep_IHS, RenewShare) over the full panel. The PVAR does not impose the ARDL long-run restriction and uses the pooled dimension of the data to recover degrees of freedom. The results mirror Table 4 qualitatively. The lagged coefficient of MonShock in the ImportDep_IHS equation is 0.063 (not significantly different from zero given a small effective R² contribution of the MonShock lag), while the lagged coefficient in the RenewShare equation is −0.001. Neither equation identifies monetary shocks as a quantitatively important driver of energy security dynamics. The autoregressive coefficients on the dependent variables are very close to unity (0.955 for ImportDep_IHS, 0.959 for RenewShare), consistent with the I(1) classification in Table 3b and with the slow error-correction speeds in Tables 4, 5. We report full PVAR output in the Supplementary material.

5 Discussion

5.1 A more conservative reading of the monetary–energy security nexus

Our central finding is that the apparent monetary policy-energy security link documented in earlier work is fragile to methodological choices that the panel time-series literature has long recommended for macro-panels of this dimension. Under the fully augmented specification (five-variable SVAR with REER endogenous, IHS transformation of ImportDep, PMG with MonShock × DeficitDummy interaction, SubsidyProxy control, CIPS unit-root tests, and robust standard errors), we cannot reject the null that monetary shocks have no long-run effect on either energy import dependence or renewable energy share. This message is more conservative than the positive findings of Wang et al. (2022), and it is important to state it clearly rather than obscure it with selective reporting.

Our null finding is not without precedent in the broader literature. Hashmi et al. (2022) document significant effects of monetary policy on renewable energy consumption in the US, but their sample involves a single advanced economy with deep capital markets and a well-developed renewable energy sector. The contrast with our ASEAN results is consistent with the hypothesis that monetary transmission to energy outcomes requires a threshold level of financial market depth and institutional capacity that many ASEAN economies have not yet reached. Similarly, while Rao et al. (2024) find that renewable energy firms in Asia are more vulnerable to interest-rate volatility than fossil fuel firms, their firm-level analysis operates at a different level of aggregation from our macro-panel approach. The firm-level sensitivity they document may not translate into detectable macro-level shifts in renewable energy share because ASEAN renewable energy deployment is driven more by government procurement mandates and feed-in tariffs than by market-based investment decisions. Dagar et al. (2024) find that economic policy uncertainty reduces energy diversification, but their measure of uncertainty is broader than our SVAR-identified monetary shocks, and their panel spans 26 countries across three continents with considerably more variation in institutional quality.

That said, three pieces of evidence suggest that a weak demand-side channel may still be present. First, the MonShock × DeficitDummy interaction is negatively signed across all estimators (PMG, MG, and CS-ARDL), which is the direction predicted by the demand-channel hypothesis. Second, the error-correction term is negative and economically plausible in all specifications, indicating that a long-run equilibrium does exist between monetary conditions and import dependence, even if its identification with the current sample size is weak. Third, the Serebriakova et al. (2026) technology-bias finding for Europe provides a mechanism through which a null effect on RenewShare is economically coherent: if contractionary policy simultaneously reduces demand (which would lower ImportDep) and depresses renewable capacity additions (which would raise ImportDep by crowding out domestic supply), the two effects could net out to the statistically indistinguishable-from-zero estimates we observe. This mechanism offers a different reading of our null finding. The absence of a significant aggregate effect does not necessarily imply that monetary policy is neutral with respect to energy security; it may instead reflect two opposing channels that cancel out in the aggregate. However, we emphasize that this interpretation remains speculative in the absence of statistically significant coefficients, and we cannot distinguish between a true null effect and an offsetting-channels explanation with the available data.

5.2 Structural heterogeneity and the interaction specification

Replacing the earlier split-sample MG estimation with a pooled PMG specification that includes a MonShock × DeficitDummy interaction is not merely a matter of statistical convenience. Split-sample MG on N = 3 and N = 4 sub-panels would yield long-run coefficients estimated from cross-country averages of three and four observations respectively, with near-degenerate standard errors and no formal basis for comparing the two sub-groups. We choose the interaction specification for three reasons. It preserves the full N = 7 sample, it delivers a direct statistical test of the deficit–surplus differential in a single regression, and it is the standard practice in the panel ECM literature for questions of this type. The magnitudes of the interaction coefficients, though individually insignificant, remain informative. They are consistently negative and comparable in size to the main effect, supporting the qualitative reading that the demand channel is stronger in deficit economies.

The interpretation of the implied long-run effects is instructive. Under PMG, the surplus-country long-run coefficient is 2.502 (driven by Brunei, Indonesia, and Malaysia) while the deficit-country coefficient is 0.360. The large surplus-country point estimate should not be taken at face value. With only three countries in the surplus group, the cross-country variance is dominated by Brunei's idiosyncratic trajectory, from a near-monoexporter to a slightly more diversified exporter over the sample. The standard error of 2.084 properly reflects this uncertainty. The deficit-country point estimate of 0.360 is much closer to the null, but its standard error is also wide, and we cannot reject either a zero or a modestly negative effect.

5.3 The energy security-energy transition dilemma revisited

Monetary tightening in principle offers mixed consequences for energy security. In the short run, higher rates can reduce import dependence by dampening aggregate demand for imported hydrocarbons. In the long run, the same rates can obstruct the investment pipeline for renewable energy, slowing the structural shift that would deliver genuinely sustainable security improvements. Our results are consistent with this dilemma but cannot determine its quantitative balance in ASEAN: the ImportDep equation shows no significant demand-side effect, the RenewShare equation shows no significant investment-side effect, and the error-correction speeds are slow enough that meaningful adjustment requires a decade or more.

The insensitivity of RenewShare to monetary conditions in ASEAN deserves deeper examination. Several structural features of ASEAN renewable energy markets may explain why monetary policy shocks do not translate into detectable changes in the renewable energy share. First, renewable energy deployment in most ASEAN economies is driven primarily by government procurement mechanisms, feed-in tariffs, and power-purchase agreements rather than by market-based investment decisions that are directly sensitive to interest rates. In Indonesia, geothermal development is governed by centralized procurement through PLN (the state utility), while in Vietnam, the rapid expansion of solar capacity in 2019–2021 was driven by time-limited feed-in tariff programs rather than by prevailing interest rates. Second, the renewable energy sector in ASEAN relies heavily on concessional financing from multilateral development banks and bilateral development finance institutions, which are insulated from domestic monetary conditions. The Asian Development Bank's Energy Transition Mechanism and the Green Climate Fund's concessional lending windows provide financing on terms that are largely independent of domestic policy rates. Third, the small absolute size of the modern renewable energy sector in most ASEAN economies (RenewShare averages only 8% in our sample) means that even substantial percentage changes in renewable investment may not register as statistically significant movements in the aggregate renewable energy share. Fourth, regulatory and grid-connection barriers, rather than financing costs, are the binding constraints on renewable deployment in several ASEAN markets (ACE, 2024). These structural factors collectively suggest that the monetary policy-renewable energy channel documented in advanced economies by Serebriakova et al. (2026) and Hashmi et al. (2022) does not operate with the same force in ASEAN, where institutional and regulatory factors dominate the investment-channel effects.

The implication for central banks and energy ministries is that conventional monetary policy should not be regarded as a fine-tuning instrument for energy security objectives in the short or medium run. The levers that move the energy mix over a five-to-ten-year horizon are fiscal subsidy reform, targeted renewable financing facilities (such as the two-tier structures discussed in the European context by Ferrari and Nispi Landi, 2023), grid-integration investment, and regulatory sequencing of power-purchase agreements.

5.4 Policy implications

Given that our empirical results do not establish a statistically significant causal link between monetary policy and energy security in ASEAN, the policy implications we draw are necessarily grounded in the theoretical framework and the supporting evidence from other regions, rather than in our own regression coefficients. We frame them as institutional design recommendations that are consistent with the theoretical channels identified in Section 2.2, while acknowledging that their empirical justification in the ASEAN context remains to be established with better data.

First, ASEAN central banks should recognize that conventional interest-rate policy is a blunt instrument for energy security objectives. Our results indicate that neither energy import dependence nor renewable energy share responds significantly to monetary shocks at conventional significance levels. This means that central banks should not expect monetary tightening to deliver energy security co-benefits, nor should they factor energy security considerations into routine interest-rate decisions. The primary implication is negative: monetary policy should not be burdened with energy security objectives that it cannot reliably achieve.

Second, if policymakers wish to insulate renewable energy investment from the effects of monetary tightening cycles, targeted financing instruments are more promising than adjustments to the headline policy rate. Concrete examples within ASEAN include: (i) Indonesia's Geothermal Fund could be expanded to cover a broader range of renewable technologies, with lending rates decoupled from the Bank Indonesia policy rate; (ii) the Philippines' Renewable Energy Trust Fund could be recapitalized and its disbursement mechanisms streamlined to provide counter-cyclical support during periods of monetary tightening; (iii) Vietnam's State Bank could establish a dedicated green refinancing window, similar to the Bank of Japan's Climate Response Financing Operations, that provides preferential-rate funding to commercial banks for renewable energy lending. These instruments would address the technology-bias channel documented by Serebriakova et al. (2026) without compromising the headline monetary policy stance.

Third, closer coordination between monetary authorities and energy regulators is warranted during monetary tightening episodes. Practical coordination mechanisms could include: (i) regular joint meetings between central bank monetary policy committees and energy regulatory agencies to assess the impact of interest-rate decisions on energy investment pipelines; (ii) pre-announced schedules for renewable energy procurement auctions that are timed to coincide with anticipated easing cycles; and (iii) temporary credit guarantee schemes for renewable energy projects during periods of elevated interest rates, funded through fiscal rather than monetary channels. The ASEAN Power Grid initiative and the ASEAN Interconnection Master Plan provide existing institutional platforms through which such coordination could be operationalized.

Fourth, fuel subsidy reform remains the most direct policy lever for strengthening the demand-side channel of monetary transmission to energy security. Our SubsidyProxy coefficients, while not statistically significant, are consistently positive in sign, suggesting that subsidies may weaken the demand-response mechanism. Indonesia's 2015 subsidy removal and Malaysia's ongoing subsidy rationalization program provide natural experiments through which this hypothesis could be tested more rigorously as post-reform data accumulate. Panda and Mahalik (2026) provide complementary evidence that governance quality mediates the relationship between renewable energy and energy uncertainty, suggesting that subsidy reform is most effective when accompanied by improvements in energy governance.

We emphasize, however, that these recommendations flow from the theoretical framework and the supporting international evidence rather than from statistically significant coefficients in our own regressions; the empirical case for large-scale coordinated intervention remains to be established with better data.

5.5 Limitations

Several limitations should be acknowledged. First, the exclusion of Cambodia, Lao PDR, and Myanmar due to data constraints reduces the sample to seven countries, which limits generalizability to the full ASEAN bloc and constrains the degrees of freedom available for heterogeneous estimators. A future extension of the dataset using IEA full balances or ASEAN Centre for Energy (ACE) primary statistics would address this limitation. Second, the absence of a continuous HHI series prevents us from testing whether monetary shocks affect the detailed composition of the energy mix beyond the binary fossil/renewable split captured by RenewShare. Third, the binary SubsidyProxy is a coarse approximation of fuel-subsidy regime heterogeneity; the development of a continuous country-year series using IMF fossil-fuel subsidy estimates or IEA subsidy data would enable a formal interaction test of the subsidy-moderation hypothesis. Fourth, the N = 7 sample dimension yields limited statistical power for detecting heterogeneous effects, as reflected in the wide confidence intervals on all key estimates. Fifth, our RenewShare series excludes traditional biomass to focus on modern renewable capacity. An alternative series that included traditional biomass might yield different conclusions for countries with large informal biomass sectors. However, such a specification would conflate the energy transition with a conceptually distinct informal-sector shift.

6 Conclusion

This study provides empirical evidence on the relationship between monetary policy and energy security outcomes in seven Southeast Asian economies over 2000–2023. We extract exogenous monetary policy shocks via country-level SVAR models in which the real effective exchange rate is endogenous, apply an inverse hyperbolic sine transformation to the import-dependence series to mitigate outlier influence, and estimate a Pooled Mean Group panel ECM with a MonShock × DeficitDummy interaction term in place of statistically underpowered split-sample estimation. We complement these main results with Mean Group, CS-ARDL, and Panel VAR robustness checks and implement second-generation CIPS unit-root tests that are robust to cross-sectional dependence.

Under this augmented specification, the long-run effect of monetary shocks on ImportDep_IHS is not statistically significant at conventional levels. The interaction with the deficit dummy carries the theoretically expected negative sign but is also not significant. Renewable energy share exhibits no response to monetary conditions under any estimator. The error-correction term is negative and of plausible magnitude but implies a half-life of approximately 9 years under PMG, indicating that any adjustment toward equilibrium is gradual. These findings temper the more optimistic message of prior work in this area and suggest that conventional monetary policy is not a reliable instrument for energy security objectives in the ASEAN context.

Our contributions are methodological as well as substantive. We show that five methodological choices, taken together, deliver a more conservative and defensible reading of the monetary–energy security nexus than the earlier literature provides. They are: SVAR-based identification with an endogenous exchange rate, IHS transformation of outlier-laden dependent variables, interaction-based heterogeneity modeling in preference to split-sample estimation, CIPS unit-root testing, and explicit treatment of cross-sectional dependence. In addition, we provide a systematic panel diagnostic sequence (cross-sectional dependence, slope heterogeneity, unit roots, and cointegration) that disciplines the choice of estimator and ensures that the estimation strategy is data-driven rather than arbitrary.

Future research should extend the sample as comprehensive data become available for Cambodia, Lao PDR, and Myanmar; incorporate continuous fossil-fuel subsidy data to test the subsidy-moderation hypothesis formally; and investigate whether unconventional monetary policies, such as green lending facilities and two-tier rate structures, offer pathways to reconcile energy security with the long-run energy transition.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frevc.2026.1855951/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Fengtao Guang, China University of Geosciences Wuhan, China

Reviewed by: Tri Kurniawati, Padang State University, Indonesia

Himanshu Sekhar Panda, Symbiosis Institute of Business Management, India